class DummyTrainer(Trainer):
def _finish_init(self):
self.net = DistFastNet(self.learning_rate, self.image_shape,
self.n_out, self.init_model)
def get_from_master(self, data):
data = comm.bcast(data, root=0)
comm.barrier()
return data
def train(self):
util.log('rank %d starting training...' % rank)
while self.should_continue_training():
train_data = self.get_from_master(None)
self.curr_epoch = self.train_data.epoch
input, label = train_data.data, train_data.labels
self.net.train_batch(input, label)
self.curr_batch += 1
if self.check_test_data():
self.get_test_error()
if self.factor != 1.0 and self.check_adjust_lr():
self.adjust_lr()
def get_test_error(self):
test_data = self.get_from_master(None)
input, label = test_data.data, test_data.labels
self.net.train_batch(input, label, TEST)
def save_checkpoint(self):
self.net.get_dumped_layers()
class DummyTrainer(Trainer):
def _finish_init(self):
self.net = DistFastNet(self.learning_rate, self.image_shape, self.n_out, self.init_model)
def get_from_master(self, data):
data = comm.bcast(data, root = 0)
comm.barrier()
return data
def train(self):
util.log('rank %d starting training...' % rank)
while self.should_continue_training():
train_data = self.get_from_master(None)
self.curr_epoch = self.train_data.epoch
input, label = train_data.data, train_data.labels
self.net.train_batch(input, label)
self.curr_batch += 1
if self.check_test_data():
self.get_test_error()
if self.factor != 1.0 and self.check_adjust_lr():
self.adjust_lr()
def get_test_error(self):
test_data = self.get_from_master(None)
input, label = test_data.data, test_data.labels
self.net.train_batch(input, label, TEST)
def save_checkpoint(self):
self.net.get_dumped_layers()
